Geometry Distributions

TL;DR

This paper introduces a novel geometric data representation using diffusion models to learn surface point distributions, improving flexibility and accuracy in modeling complex 3D structures.

Contribution

It proposes a distribution-based geometric representation with a new neural architecture, addressing limitations of existing coordinate-based methods.

Findings

Achieves high geometric fidelity across various object types

Effective in textured mesh representation and neural surface compression

Enhances dynamic object modeling and rendering capabilities

Abstract

Neural representations of 3D data have been widely adopted across various applications, particularly in recent work leveraging coordinate-based networks to model scalar or vector fields. However, these approaches face inherent challenges, such as handling thin structures and non-watertight geometries, which limit their flexibility and accuracy. In contrast, we propose a novel geometric data representation that models geometry as distributions-a powerful representation that makes no assumptions about surface genus, connectivity, or boundary conditions. Our approach uses diffusion models with a novel network architecture to learn surface point distributions, capturing fine-grained geometric details. We evaluate our representation qualitatively and quantitatively across various object types, demonstrating its effectiveness in achieving high geometric fidelity. Additionally, we explore applications using our representation, such as textured mesh representation, neural surface compression, dynamic object modeling, and rendering, highlighting its potential to advance 3D geometric learning.